Need Jibo's ASR/TTS in the simulator

The text/based ASR/TTS simulation in the SDK is nice, but it is not enough.

In order to really understand and test user speech and Jibo response, it is necessary to get access to live speech recognition (ASR) and to hear Jibo’s output (TTS). Typed text input will not be subject to speech recognition errors that we will need to handle. And the cadence of user interaction cannot be adequately appreciated and tuned without real-time spoken interaction.

I have been able to fake it by setting up VAD (Voice Activity Detection) from npm libraries, then calling out to Nuance ASR and TTS myself. But this is slow and klunky (at least I hope it is slower than Jibo’s native calls to ASR and TTS!). And the only voices available to non-paying Nuance developer accounts are Samantha and Tom. Samantha is female, and Tom is the voice of the ubiquitous and extremely annoying Tom Glynn—nothing like Jibo.

Please make Jibo’s ASR and TTS available in the simulator.

4 Likes

Or provide us with Jibo units we can test on ourselves! :smiley:

2 Likes

Seconded, @alfarmer. :smile: The simulator is very robust, but as you probably already know internally, testing on the hardware is critical for finalizing a skill, especially those that have strong interaction between the user and robot. For instance, I’m in the process of building a full-fledged kitchen assistant with guided recipes and it’s near impossible to test this skill in a real-world setting without the robot itself.

3 Likes

I too have the same issue with my Jibo Presentation Manager. Testing is near impossible without the Robot; as the spacial, timing of multi-person interactions can’t be simulated :frowning:

This is why I don’t start to develop with Jibo… it needs that the simulator can also simulate the HW (Speech, etc…) or (the best) is to have Jibo HW in order to develop in a better way…

I will wait until October… then I will decide about it, it depends from the market that is moving very fast!